http://www.baka-tsuki.org/project/index.php?action=history&feed=atom&title=User%3AZiggr 2026-05-02T20:18:49Z Revision history for this page on the wiki MediaWiki 1.43.1 http://www.baka-tsuki.org/project/index.php?title=User:Ziggr&diff=527964&oldid=prev 2017-09-30T05:28:27Z

<p>Gokigenyou. Ziggr desu.</p> <p><b>New page</b></p><div>=Gokigenyou.=<br /> <br /> Ziggr desu.<br /> <br /> Ten years ago, I translated the first volume of [[Maria-sama ga Miteru]] as an educational project. It was &#039;&#039;way&#039;&#039; above my reading level. The translation is pretty awful. Thank goodness for Seki&#039;s much better translation here.<br /> <br /> Ten years later, my Japanese language skill has improved little, but I&#039;m slogging through it.<br /> <br /> =Tool Chain=<br /> <br /> # &#039;&#039;&#039;Purchase book.&#039;&#039;&#039; Local Kinokuniya bookstore, or online via amazon.co.jp<br /> # &#039;&#039;&#039;Rip pages out of book.&#039;&#039;&#039; Clean up with a paper cutter.<br /> # &#039;&#039;&#039;Scan pages.&#039;&#039;&#039; Fujiscan sheet-feeding scanner can scan a whole book in just a few minutes.<br /> # &#039;&#039;&#039;OCR pages.&#039;&#039;&#039; I use an ancient copy of ReadIris. It does a remarkably great job with most characters and an infuriatingly bad job with a few.<br /> # &#039;&#039;&#039;Correct OCR errors.&#039;&#039;&#039; This takes almost as long as translation. But it&#039;s worth it, because the fewer the errors in the source text, the easier it is to get useful machine translation and dictionary lookups. I tend to correct 10-20 pages in a batch. After this, my eyes glaze over and I stop seeing errors.<br /> # &#039;&#039;&#039;Translate.&#039;&#039;&#039; I use OmegaT for the UI. Its ability to re-use translations is not particularly helpful for novels of full sentences, but I like having &#039;&#039;some&#039;&#039; sort of UI that can show Japanese and English next to each other as I work through each page. I set up Google Translate API so that I get automatic machine translations. These are almost always hilariously bad, but can often pick up a couple meanings that dictionaries miss. I use jisho.org for my dictionary.<br /> # &#039;&#039;&#039;Edit for grammar and smoothness.&#039;&#039;&#039; Sentence-by-sentence translation in OmegaT produces a stumbling translation that needs a bit more work. So I paste each translated chapter into Grammarly.com and read it as a chapter. Fix typos and grammar errors. Make things read a bit more smoothly.<br /> <br /> =Translation Stylesheet=<br /> <br /> &#039;&#039;&#039;Retain iconic phrases.&#039;&#039;&#039; Gokigenyou, onee-sama, and honorifics are all key identifying phrases. Leave them unchanged.<br /> <br /> &#039;&#039;&#039;Translate meaning, not words.&#039;&#039;&#039; The Japanese text includes many amazingly complex compound sentences that are nigh unreadable if transliterated to English. It&#039;s okay to break up paragraph-long compound sentences. It&#039;s okay to insert speaker names when it might be unclear. It is okay to omit some words. <br /> <br /> One thing I lack the ability to do, but really wish that I could, is to retain the flavor of each speaker. Yoshino is a bit more brusque than Yumi. Shimako is soft and prone to kashira.<br /> <br /> [[User:Ziggr|Ziggr]] ([[User talk:Ziggr|talk]]) 07:28, 30 September 2017 (CEST)</div>

Ziggr